16
$\begingroup$

I understand that each kernel implies a particular feature map. For instance for $x,z \in R^2$ the kernel $K(x,z)=(\textrm{dot}(x,z))^2$ implies a feature map $$\langle\phi(x_1), \phi(x_2)\rangle=\langle [x_1^2 , x_1 x_2 , x_1 x_2, x_2^2], [z_1^2 , z_1 z_2 , z_1 z_2, z_2^2] \rangle$$, as though my inputs are transformed from $R^2$ to $R^4$ in the process of computing $K(x,z)$.

The polynomial kernels are clear enough to see in this way. However the radial basis function kernel in particular, i.e. $K(x,z) = \exp(-\|x-z\|^2/c^2)$, supposedly implies an infinite dimensional feature map. Exactly why is this so?

$\endgroup$
16
$\begingroup$

Try expanding the exponential into an infinite series, and you'll see it. Abu-Mostafa has a great demonstration of this, in his classic style.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.